Stanford researchers have developed an AI system, Tutor CoPilot, using OpenAI's GPT-4.
Today, education has become unbelievably expensive, making it a nightmare for many families. In Tier 1 and Tier 2 Indian cities, the monthly tuition fees for private schools often ranges between INR 2,500 to INR 10,000. This, of course, is in addition to fees towards maintenance, lab and technology, books and supplies, and transportation.
The global requirement for teachers adds another layer to this challenge. A recent UNESCO report highlights that by 2030, the world will need 44 million new primary and secondary teachers to meet the global education goals.
The financial implications are equally concerning. Achieving universal education goals will require an additional $ 12.8 billion for primary education and $ 106.8 billion for secondary education every year. In total, funding these new teaching positions demands $120 billion annually by 2030, stressing the urgent need for global investment in education.
So, how do we solve this?
Stanford researchers have developed an AI system, Tutor CoPilot, using OpenAI's GPT-4, and integrated it into FEV Tutor, a platform that virtually connects students with tutors. This human-AI collaboration aims to provide tutors with "expert-like" support, enhancing their teaching by offering guidance that mirrors an expert's thinking.
The study, a randomised controlled trial in live tutoring, involved 900 tutors and 1,800 K-12 students from historically underserved communities. The results noted that the students working on maths with tutors equipped with Tutor CoPilot were four percentage points more likely to master topics compared to those in the control group - a statistically significant improvement (p<0.01).
Students paired with lower-rated tutors saw even greater gains, with topic mastery rising by nine percentage points.
Beyond effectiveness, Tutor CoPilot is also affordable, costing just $20 per tutor annually.
Further, analysing over 5,50,000 tutor-student messages, researchers found that tutors with Tutor CoPilot were more likely to use pedagogical strategies that deepened student understanding, such as asking guiding questions, rather than simply giving away answers.
In the UK, TLC LIVE recently introduced Manda, an AI tutor aligned with the national curriculum to support Key Stage 3 and 4 students in maths and English. Developed on Meta's Llama 3 model, Manda draws from the collective expertise of 300 qualified UK teachers, providing affordable AI tutoring for just £10 a month.
Like Tutor CoPilot, Manda brings the expertise of seasoned educators to every session, ensuring that students across the country have access to dependable academic support.
In China, the potential of AI in education is being leveraged to close the gap between urban and rural educational resources. At Baishaping Primary School, where many students had been struggling in maths and literacy, an adaptive learning platform powered by Squirrel Ai Learning was implemented.
This system tailors lessons based on each student's unique needs, creating customised pathways for learning. This adaptive approach is helping students in rural areas receive the same quality of education as their urban counterparts, reducing regional disparities in access to high-quality instruction.
At the recent Build with AI Summit in Bengaluru, Nandan Nilekani, the co-founder and non-executive chairman of Infosys, said, "In India, billions of people can't read. Now, if you can use AI to improve the reading skills of every child, then they will also join the reading."
It's an ambition that aligns with the growing wave of AI-driven educational tools emerging in India, aimed at levelling the playing field for students across the nation.
Leading the charge is Wadhwani AI, a nonprofit organisation. One of their tools is designed to assess reading accuracy, speed, and errors, providing teachers with insights into each child's oral reading fluency.
By analysing audio clips of students reading, this AI model can highlight areas of strength and pinpoint where additional support may be needed, contributing to better reading comprehension and overall learning outcomes.
Wadhwani AI is also tackling the issue of student dropouts, an area where India lacks standardised interventions. Their dropout prediction tool identifies students at risk of leaving school based on various social and academic factors. This AI model acts as an early warning system, enabling educators to step in with timely, targeted support to keep students engaged and enrolled.
While India hasn't yet introduced AI tutors on a large scale, the country is making a headway with Iris, India's first AI-powered teacher robot. Deployed at KTCT Higher Secondary School in Thiruvananthapuram, Iris is a unique robot in a saree, created in collaboration with Makerlabs Edutech.
This AI teacher was built under the Atal Tinkering Lab (ATL) project by NITI Aayog, aiming to bring innovative, personalised learning experiences to students.
In May, Khan Academy partnered with Microsoft to develop Phi-3, a specialised AI maths tutor, while OpenAI introduced ChatGPT Edu, tailored for universities to responsibly integrate AI across classrooms, research labs, and administrative teams.
In India, the edtech startup Physics Wallah launched Alakh AI, featuring AI Guru -- a 24/7 personal tutor designed to support students beyond academics. From answering subject-specific questions to assisting with product and support issues, AI Guru delivers responses in both text and video formats, adapting to each student's unique learning needs.
As the future of AI in education is unfolding quickly, AI tutors could soon become a reliable teaching resource for students, especially in rural and underserved regions.
